Actual limitations on scan_interval?

I’m over simplifying this a bit, but every state change event is sent to:

  • all frontend clients
  • mobile apps
  • the database (and stored on disk)
  • integrations that listen for all state changes

If you poll anything too frequently you can end up with

  • sluggish frontend
  • battery drain on mobile
  • bloated database
  • downline performance issues with listening integrations

You can throw more resources (cpu, bigger batteries, ram, faster disks, etc) at the above to reduce the effects but in general integrations that need more frequent updates than five seconds should be push and filter out duplicate or useless data.

Personally, I usually replace devices that require frequent polling with something that can push state in lieu of upgrading my ha/mobile/desktop hardware as I find it’s more cost effective and reliable long term